3 days of thinking. 6 hours of building.

I audited 14 Borivali elder-care sites. 90% were broken templates or generic stock; the one or two that ranked still felt institutional — clinical, dated, cold. The design-quality bar in the local market was completely undefended.

The 18-resident cap wasn't a weakness. It was the product.

The obvious read was scale: a small home with 18 beds can't compete with multi-building campuses on amenities, budget, or capacity — so compete on price or polish. That's the trap. Four hours of discovery with the founder reframed it.

The deeper insight was the actual user. The person choosing isn't the resident — it's an adult child, 40–65, making one of the most emotionally charged decisions of their life in a moment of acute concern. They aren't browsing; they're frightened. The site's job is to reduce that anxiety fast enough that they pick up the phone. Positioning followed from one sentence — “A home, not a facility” — and everything downstream fell out of it: visual direction, photography brief, content tone, the hierarchy of trust signals.

One positioning idea carried across every surface: warm photography over claims, real names and faces, dense trust signals, identical templates so 45+ pages feel like one product.

Homepage — trust signals first.

Govt-registration line up top, real photography (not stock), two-tap CTAs (Schedule a Visit / Call). The photography brief was given to the founder so every image was a real moment, not a stock library.

Services — 12 care types, each a pillar SEO page.

Senior Citizen, Alzheimer's & Dementia, Paralytic & Stroke, Cancer Care, Post-Surgery Recovery, NRI Elder Care, and six more. Each card maps to a pillar page from an identical template — same heading hierarchy, checklist density, and CTA placement, enforced by a Claude Code skill and repeatable across all 12.

Daily routine — the question adult children actually ask.

The most-asked question is “what does a day look like?” The schedule answers it before they ask — morning tea through night rest — with a footnote acknowledging medical adjustments. Honesty over polish.

Our Story — trust is built by people, not logos.

Founders Sammeer and Ssunita Nawathe on the page with names, faces, and why they started. “Aannapurnaa” is the Goddess of nourishment; “Aai” means mother in Marathi. Trust on an emotional decision is built by people, not logos.

Facilities — the room a parent will sleep in.

Shown, not described. Adult children compare physical environment first, copy second.

Articles — the visible surface of the pillar-cluster system.

30+ in-depth guides indexed by category, each with an English / Hindi / Marathi toggle, a Key Takeaways block, on-this-page nav, and dense internal linking back to the pillar service page it supports.

Speed is a byproduct of clarity. The 6-hour build was only possible because three days of thinking had already answered every question Claude would otherwise have resolved mid-execution. Of 78 hours total, 72 were thinking and 6 were building.

Blog content was research-first: for every topic, Claude queried DataForSEO for ranking keywords, search volume, and competitor content, then wrote articles built to rank from the first word — not blogs tagged with keywords after writing. Each of the 12 services became a pillar with 2–3 cluster articles linking back to it and across to related clusters.

Eight decisions that did the disproportionate work:

The skills are the deliverable.I wrote four SKILL.md files before any production code — design-system, singleton-pages, service-detail-pages, and blog-pages. Each exists because I identified a failure pattern Claude would hit at scale and hardcoded the fix. They're quality-control systems, not shortcuts.

The Hour-2 gate.At Hour 1, I had Claude generate two throwaway pages from the skills as written. At Hour 2, I reviewed them line by line. Three failure patterns appeared — tone drift, shallow research (drafting without forcing a DataForSEO lookup), and layout drift — and each was a gap in the skills, not in Claude. Every fix went back into the relevant skill before any production page was generated. Consistency after the fixes was ~95% across all 45+ pages, with no per-page review cycle. The gate isn't a code review; it's a skill review.

Three months of Search Console from a brand-new domain: 28 clicks, 1.41K impressions, average position 5.8 — page 1 for brand terms and several local-intent terms, the compounding curve a pillar-cluster build is designed to produce. 124 unique ranking queries, brand plus generic intent (“old-age home in Borivali West,” “best old-age home in Mumbai”) — the cluster strategy capturing both.

97 / 94 / 100 / 100 with LCP 1.1s on a mobile-first audience. Engineering quality matches design quality — Bootstrap + static HTML wasn't the lazy choice, it was the right one.

What this taught me about AI-assisted development.The mistake most professionals make is treating Claude Code as a faster typist. It isn't — it's a junior engineer that executes literally whatever it's asked, and at scale that produces inconsistent output. The skills are how you stop being a prompter and start being an architect: write the skills first, run an Hour-2 gate to find the gaps, patch the skills, then let the agent ship at scale.

Two honest notes for next time:analytics should have been wired before launch, not after — the result is confirmed directionally (the founder says families are calling) but I can't quantify it as precisely as I'd like; and a lightweight CMS (even a Notion-to-HTML pipeline) would let the founder make basic updates without me — the current setup works because I'm available, which is a dependency, not a solution.